Switching costs and network effects
in competition policy
Jacques Crémer
Toulouse School of Economics
with credit to
Gary Biglaiser
University of North Carolina
and
Gergely Dobos
Gazdasági Versenyhivatal (GVH)
CREESE, Rhodes, Greece
2 July 2011
Levin on Internet markets
“In traditional industries with network effects, high switching
costs are often an important compounding factor.
Levin on Internet markets
“In traditional industries with network effects, high switching
costs are often an important compounding factor. Consider the
case of operating systems, where switching costs can be
relatively high for individual users and for firms with large
computer installations.
Levin on Internet markets
“In traditional industries with network effects, high switching
costs are often an important compounding factor. Consider the
case of operating systems, where switching costs can be
relatively high for individual users and for firms with large
computer installations. Switching between internet platforms
or using multiple platforms can be considerably easier. That is,
one can shop on Amazon and eBay, or be a Facebook user and
try Twitter.
Levin on Internet markets
“In traditional industries with network effects, high switching
costs are often an important compounding factor. Consider the
case of operating systems, where switching costs can be
relatively high for individual users and for firms with large
computer installations. Switching between internet platforms or
using multiple platforms can be considerably easier. That is,
one can shop on Amazon and eBay, or be a Facebook user and
try Twitter. At least in some cases, the combination of low
switching costs and low costs to creating new platforms might
mitigate traditional concerns about lock-in and dynamic
inefficiency.”
Levin on Internet markets
“In traditional industries with network effects, high switching
costs are often an important compounding factor. Consider the
case of operating systems, where switching costs can be
relatively high for individual users and for firms with large
Switching
internet
computer installations.
But what do
we knowbetween
about the
rela- platforms or
using multipletionship
platforms
can beswitching
considerably
between
costseasier.
and That is,
one can shopnetwork
on Amazon
and
eBay,
or
be
a
Facebook
user and
effects?
try Twitter. At least in some cases, the combination of low
switching costs and low costs to creating new platforms might
mitigate traditional concerns about lock-in and dynamic
inefficiency.”
Three themes
How do switching cost models and network
models differ?
Three themes
How do switching cost models and network
models differ?
How do we model inertia in networks?
Three themes
How do switching cost models and network
models differ?
How do we model inertia in networks?
What are the consequences of heterogeneity of consumers in dynamic models of both
types?
Three themes
How do switching cost models and network
models differ?
How do we model inertia in networks?
What are the consequences of heterogeneity of consumers in dynamic models of both
types?
How do switching cost and network effects
mix?
Three themes
How do switching cost models and network
models differ?
How do we model inertia in networks?
What are the consequences of heterogeneity of consumers in dynamic models of both
types?
How do switching cost and network effects
mix?
No two sidedness!
The simplest possible models
+ Main assumptions:
å One incumbent;
å Free entry;
å No discrimination.
The simplest possible models
+ Main assumptions:
å One incumbent;
å Free entry;
å No discrimination.
+ Switching cost
price = profit = σ.
å Efficient
The simplest possible models
+ Main assumptions:
å One incumbent;
å Free entry;
å No discrimination.
+ Switching cost
price = profit = σ.
å Efficient
+ Network effects
price = profit = ν.
å Efficient.
Modeling coordination failures in network effects
Modeling coordination failures in network effects
+ Crémer, Rey, Tirole: a mass of “trapped” consumers.
Modeling coordination failures in network effects
+ Crémer, Rey, Tirole: a mass of “trapped” consumers.
+ Caillaud, Jullien: coordination on worse equilibrium for the
entrant.
Modeling coordination failures in network effects
+ Crémer, Rey, Tirole: a mass of “trapped” consumers.
+ Caillaud, Jullien: coordination on worse equilibrium for the
entrant.
+ Trying to say things about the whole set of equilibria.
Modeling coordination failures in network effects
+ Crémer, Rey, Tirole: a mass of “trapped” consumers.
+ Caillaud, Jullien: coordination on worse equilibrium for the
entrant.
+ Trying to say things about the whole set of equilibria.
+ Our solution: strong non-coordination.
Efficiency issues revisited
Assume that the entrants offer (stand-alone) utility W + ε.
+ With switching costs: consumers pay σ − ε. Efficiency is
preserved.
Efficiency issues revisited
Assume that the entrants offer (stand-alone) utility W + ε.
+ With switching costs: consumers pay σ − ε. Efficiency is
preserved.
+ With network effects: consumers pay ν − ε. Inefficient
equilibrium.
Efficiency issues revisited
Assume that the entrants offer (stand-alone) utility W + ε.
+ With switching costs: consumers pay σ − ε. Efficiency is
preserved.
+ With network effects: consumers pay ν − ε. Inefficient
equilibrium.
å Note that we can have inefficiency without discrimination.
Elementary repeated games with homogenous
consumers
p = (−δσ + σ) + δσ = σ.
Elementary repeated games with homogenous
consumers
p = (−δσ + σ) + δσ = σ.
p = (−δν + ν) + δν = ν.
Elementary repeated games with homogenous
consumers
p = (−δσ + σ) + δσ = σ.
p = (−δν + ν) + δν = ν.
You do not become rich on switching costs (or network effects) alone.
Heterogeneity of consumers: static model
Some consumers with switching costs equal to zero (or with no
value for network effect).
price = σ;
profit = ασ.
Heterogeneity of consumers: static model
Some consumers with switching costs equal to zero (or with no
value for network effect).
price = σ;
profit = ασ.
price = αν
profit = α2 ν.
Heterogeneity of consumers: static model
Some consumers with switching costs equal to zero (or with no
value for network effect).
price = σ;
profit = ασ.
price = αν
profit = α2 ν.
Remark: With heterogeneous consumers a no
discrimination rule can be costly in terms of social
welfare
Dynamics
with
heterogeneous consumers
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ + σ) + δΠ
ασ
.
=⇒ Π =
1 + αδ − δ
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ + σ) + δΠ
ασ
.
=⇒ Π =
1 + αδ − δ
Profit is greater than one
period profit . . .
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ + σ) + δΠ
ασ
.
=⇒ Π =
1 + αδ − δ
Profit is smaller than
discounted flow of one
period profit.
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ + σ) + δΠ
ασ
.
=⇒ Π =
1 + αδ − δ
Adding zero switching
cost customers increase
the profit of the incumbent.
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ + σ) + δΠ
ασ
.
=⇒ Π =
1 + αδ − δ
When δ → 1, Π → σ.
With an ∞ horizon, the profit is not equal to the one
period profit, ασ.
Π = α(−δΠ
σ) +with
δΠ
Same
results+hold
network effects. ασ
.
=⇒ Π =
1 + αδ − δ
σL > 0
is different from
νL > 0.
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
,
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
(1 − αδ)σH + δσH = (1 + δ − αδ)σH
,
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
(−δσL + σH ) + δσL = σH
< (1 − αδ)σH + δσH = (1 + δ − αδ)σH
,
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
(−δσL + σH ) + δσL = σH
< (1 − αδ)σH + δσH = (1 + δ − αδ)σH
< (−δσL + σH ) + δσH = (1 + δ)σH − δσL ,
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
(−δσL + σH ) + δσL = σH
< (1 − αδ)σH + δσH = (1 + δ − αδ)σH
< (−δσL + σH ) + δσH = (1 + δ)σH − δσL ,
a proportion strictly between 0 and 1 of σH consumers will
purchase from an entrant.
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
High switching cost customers try
(−δσL to
+ “hide”
σH ) + δσ
among
low swithching cost
L = σH
customers.
< (1 − αδ)σH + δσH = (1 + δ − αδ)σH
< (−δσL + σH ) + δσH = (1 + δ)σH − δσL ,
a proportion strictly between 0 and 1 of σH consumers will
purchase from an entrant.
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
Nothing like this happens with net(−δσL + σH ) + δσL = σH
work effects: consumers are myopic.
< (1 − αδ)σH + δσH = (1 + δ − αδ)σH
< (−δσL + σH ) + δσH = (1 + δ)σH − δσL ,
a proportion strictly between 0 and 1 of σH consumers will
purchase from an entrant.
Two periods: σL > 0 and ασH > σL
+ In 1st period,
1. incumbent charges (1 − αδ)σH (this requires some work);
2. entrants charge −δσL .
+ Because
The dynamics of erosion of market
with switch(−δσL share
+ σH ) are
+ δσvery
σH
L = different
ing<costs
effects.
(1 − and
αδ)σnetwork
H + δσH = (1 + δ − αδ)σH
< (−δσL + σH ) + δσH = (1 + δ)σH − δσL ,
a proportion strictly between 0 and 1 of σH consumers will
purchase from an entrant.
Network effects
+
switching costs
σ and ν — static
+ In static model with only network effects, incumbent
charges ν;
+ In static model with only switching costs, incumbent
charges σ.
σ and ν — static
+ In static model with only network effects, incumbent
charges ν;
+ In static model with only switching costs, incumbent
charges σ.
+ Focal equilibrium with both effects: incumbent charges
σ + ν.
+ Profits are the sum of the profits in the pure network model
and in the pure switching cost model.
More interesting
+ 1/2 consumers have switching cost 0 and 1/2 switching
cost σ. Assume also
σ < ν.
More interesting
+ 1/2 consumers have switching cost 0 and 1/2 switching
cost σ. Assume also
σ < ν.
+ With both effects present, if the incumbent charges ν + ε,
the 0 switching cost customers switch.
+ Then, the σ switching cost customers will also switch.
More interesting
+ 1/2 consumers have switching cost 0 and 1/2 switching
cost σ. Assume also
σ < ν.
+ With both effects present, if the incumbent charges ν + ε,
the 0 switching cost customers switch.
+ Then, the σ switching cost customers will also switch.
=⇒ The focal equilibrium has the incumbent charge ν.
Additivity disappears.
An illustrative story
Steve Jobs
"Some have argued that once a consumer purchases a body of
music from one of the proprietary music stores, they are forever
locked into only using music players from that one company.
Steve Jobs
"Some have argued that once a consumer purchases a body of
music from one of the proprietary music stores, they are forever
locked into only using music players from that one company.
Some have argued that once a consumer purchases a body of
music from one of the proprietary music stores, they are forever
locked into only using music players from that one company.
Steve Jobs
"Some have argued that once a consumer purchases a body of
music from one of the proprietary music stores, they are forever
locked into only using music players from that one company.
Some have argued that once a consumer purchases a body of
music from one of the proprietary music stores, they are forever
locked into only using music players from that one company.
Or, if they buy a specific player, they are locked into buying
music only from that company’s music store. . .
. . . On average, that’s 22 songs purchased from the iTunes
store for each iPod ever sold.
. . . On average, that’s 22 songs purchased from the iTunes
store for each iPod ever sold. Today’s most popular iPod holds
1000 songs, and research tells us that the average iPod is
nearly full. This means that only 22 out of 1000 songs, or under
3% of the music on the average iPod, is purchased from the
iTunes store and protected with a DRM.
. . . On average, that’s 22 songs purchased from the iTunes
store for each iPod ever sold. Today’s most popular iPod holds
1000 songs, and research tells us that the average iPod is
nearly full. This means that only 22 out of 1000 songs, or under
3% of the music on the average iPod, is purchased from the
iTunes store and protected with a DRM.
It’s hard to believe that just 3% of the music on the average iPod
is enough to lock users into buying only iPods in the future."
John Lech Johansen
“Many iPod owners have never bought anything from the iTunes
Store. Some have bought hundreds of songs. Some have
bought thousands. At the 2004 Macworld Expo, Steve revealed
that one customer had bought $29,500 worth of music.
John Lech Johansen
“Many iPod owners have never bought anything from the iTunes
Store. Some have bought hundreds of songs. Some have
bought thousands. At the 2004 Macworld Expo, Steve revealed
that one customer had bought $29,500 worth of music.
If you’ve only bought 10 songs, the lock-in is obviously not very
strong. However, if you’ve bought 100 songs ($99),
10 TV-shows ($19.90) and 5 movies ($49.95), you’ll think twice
about upgrading to a non-Apple portable player or set-top box.
John Lech Johansen
“Many iPod owners have never bought anything from the iTunes
Store. Some have bought hundreds of songs. Some have
bought thousands. At the 2004 Macworld Expo, Steve revealed
that one customer had bought $29,500 worth of music.
If you’ve only bought 10 songs, the lock-in is obviously not very
strong. However, if you’ve bought 100 songs ($99),
10 TV-shows ($19.90) and 5 movies ($49.95), you’ll think twice
about upgrading to a non-Apple portable player or set-top box.
In effect, it’s the customers who would be the most valuable to
an Apple competitor that get locked in. The kind of customers
who would spend $300 on a set-top box.”
Conclusions
+ Distribution of switching costs/network effects is important.
+ Even consumers to which the incumbent/dominant firm
does not sell can influence the outcome.
+ There are still many things we do not understand at the
fundamental theoretical level about the dynamics of
markets with switching costs and/or network effects.
+ Identifying anti-competitive behavior requires close
attention to the specific of the case.